Robust Subspace Recovery Layer for Unsupervised Anomaly Detection

Chieh-Hsin Lai, Dongmian Zou, Gilad Lerman

Code: [ dmzou/RSRAE](https://github.com/dmzou/RSRAE) + [ 1 community implementation](https://paperswithcode.com/paper/?openreview=rylb3eBtwr)

Data: [CIFAR-10](https://paperswithcode.com/dataset/cifar-10), [Caltech-101](https://paperswithcode.com/dataset/caltech-101), [Fashion-MNIST](https://paperswithcode.com/dataset/fashion-mnist), [MNIST](https://paperswithcode.com/dataset/mnist), [Reuters-21578](https://paperswithcode.com/dataset/reuters-21578), [STL-10](https://paperswithcode.com/dataset/stl-10), [cats_vs_dogs](https://paperswithcode.com/dataset/cats-vs-dogs)

TL;DR: This work proposes an autoencoder with a novel robust subspace recovery layer for unsupervised anomaly detection and demonstrates state-of-the-art results on various datasets.

Abstract: We propose a neural network for unsupervised anomaly detection with a novel robust subspace recovery layer (RSR layer). This layer seeks to extract the underlying subspace from a latent representation of the given data and removes outliers that lie away from this subspace. It is used within an autoencoder. The encoder maps the data into a latent space, from which the RSR layer extracts the subspace. The decoder then smoothly maps back the underlying subspace to a ``manifold" close to the original inliers. Inliers and outliers are distinguished according to the distances between the original and mapped positions (small for inliers and large for outliers). Extensive numerical experiments with both image and document datasets demonstrate state-of-the-art precision and recall.

Keywords: robust subspace recovery, unsupervised anomaly detection, outliers, latent space, autoencoder

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:1904.00152/code)